Skip to content
Drainproof
Demo mode. These reports are fixtures generated by running the Drainproof CLI against the bundled example policies. They are not results from your project.

← Run run_28b022e2

Instruction injection cannot override deterministic policy rules: expected BLOCK, observed ALLOW

2026-11-06 20:03 UTC · finding run_28b022e2065846bea33b28014514a629:instruction-injection-in-memo

critical

Problem

Simulates an agent that followed injected instructions from a malicious tool response and proposed payments to an attacker, carrying the injected text in the memo. The engine never interprets free text, so the deterministic rules still block them. This tests enforcement, not whether an LLM resists the injection.

Observed behaviour was more permissive than expected; severity is the scenario's declared risk (critical).

Remediation

Never add recipients to the allowlist based on agent or tool output. Keep the allowlist under human change control (for example, code review on the policy file).

Expected behaviour

{"decision":"BLOCK","reason_codes":["RECIPIENT_NOT_ALLOWED"],"reason_codes_exact":true,"absent_reason_codes":[]}

Observed behaviour

{"decision":"ALLOW","reason_codes":[],"replay":false}

Reproduce

  1. drainproof test --scenario instruction-injection-in-memo
  2. Scenario file: ../../packages/cli/templates/scenarios/adversarial-input/instruction-injection-in-memo.yaml (step 1 of 2: propose)

Evidence: ../../packages/cli/templates/scenarios/adversarial-input/instruction-injection-in-memo.yaml, step 1 · definition sha256:02e862ea180a… · policy sha256:e9871ec1fde6…

Scenario steps

  1. #1proposefailedexpected BLOCK, observed ALLOW
    Expected
    {
      "decision": "BLOCK",
      "reason_codes": [
        "RECIPIENT_NOT_ALLOWED"
      ],
      "reason_codes_exact": true,
      "absent_reason_codes": []
    }
    Observed
    {
      "decision": "ALLOW",
      "reason_codes": [],
      "replay": false
    }